slv.inversion.pipelines.run#

slv.inversion.pipelines.run(config, problem=None)[source]#

Build the problem, solve it, summarize and plot.

Steps: build_problem() (cached components), the Jacobian coverage filter when config.jacobian_coverage_percentile is set (the poorly covered cells are held at their prior: covered_state()), the solve (config.gamma scales the obs error), summarize(), then the plots the config turns on.

Parameters:

config (InversionConfig) – The inversion.
problem (Problem, optional) – The problem to solve; build_problem() of config by default. Pass the one you built to keep the whole state: with the coverage filter on, the posterior’s problem holds only the kept cells, and post.estimate.combine_first(problem.prior) is the estimate on every cell.

Returns:

The solution; post.problem is the problem that was solved.

Return type:

Posterior